Like mean, median, and standard deviation, mode as the value that appears most often in a set of data is an important feature of a distribution. The numerical value of the mode is the same as that of the mean and median in a symmetric distribution but may be very different in a highly skewed distribution. Mode regression, which models the relationship between the mode of a dependent variable and some covariates, was first introduced by Lee in terms of truncated dependent variables. Some modifications of the truncated mode regression have been proposed recently. However, little progress is made on the computation or algorithm of fitting a mode regression due to an NP-hard optimization problem. In this paper we first introduce the popular simulated annealing (SA) to solve the truncated mode regression optimization. Experiments with simulations compare favorably to SA. Then, a mode regression with the proposed algorithm is applied to explore the typical income structure of China. We also compare the income returns to gender, education, experience, job sector, and district between the majority of workers with typical income and the workers with mean, middle income via comparison between mode regression, mean regression, and median regression.
Part of the book: Computational Optimization in Engineering